Hoppa över navigering

Getting started with IoT: driving business action through analytics and automation

Inlägg på 9 augusti, 2018

Senior Program Manager, Azure IoT

Whether you’re looking for in-the-moment insights or long-term trends, the value of information is only as good as its usefulness for your business. Some information must be acted upon immediately, but it can also be stored for long-term, big-picture analysis of trends over time. For each case there are different tools to use, and Azure IoT has you covered.

Near real-time data analysis

Azure Stream Analytics is a pay-as-you-go service that ingests data from a variety of sources. In addition to real-time data from IoT Hub, it can also take in historical data from Azure Blob Storage, and combine Event Hub data with other sources to highlight correlations or run comparative analyses.

Using Azure Portal, you can initiate an Azure Stream Analytics job, direct it to the appropriate data set, and provide instructions on how to look for specific data, patterns, or relationships.

Once Stream Analytics has completed a job, you can direct the results to Blob Storage, SQL Server, a Data Lake, or a database-as-a-service offering, such as Cosmos DB. You can also send the data to HD Insights or Power BI for additional analysis and visualization.

As its name suggests, Azure Time Series Insights is designed to store, index, query, and visualize any data that is chronological or sequential in nature. And it can do so at scale, whether processing terabytes of data or billions of events.

Some of the most common applications of Time Series Insights are:

  • Conducting root-cause analysis
  • Building custom apps that analyze or visualize time series data
  • Sharing data from different locations with business leads to enable better collaboration

If you’re dealing with large data sets, Apache Spark may be your tool of choice. It uses parallel processing, making it an ideal solution for long-term analytics jobs. To improve the efficiency of its analysis it breaks data down into smaller chunks before processing, so you will likely experience some latency issues, though only a few seconds.

Historical data analysis

Looking for long-term trends can be a bit like reading tea leaves. You might consider using Azure Machine Learning (AML), an end-to-end solution for data science and advanced analytics. Using existing data sets, it can forecast future outcomes and trends. Over time, Azure Machine Learning will develop an understanding of the area of focus, providing the foundation for developing more effective experiments and gaining better outcomes.

AML also offers specialized libraries of Python code to help accelerate development of machine learning models in particular areas, including computer vision, financial and demand forecasting, and text analysis.

The original open source solution for big data analysis, Azure HD Insight is a fully managed service that is based on Hadoop. HD Insight supports several open source frameworks, making it a flexible, fast and cost-effective solution. And because it supports Kafka for HD Insight, you have the ability to create a managed, cost-effective streaming analytics solution that can scale to handle massive amounts of data.

Key benefits of HD Insights include:

  • Integration with Azure Log Analytics gives you a single portal for monitoring the performance of your clusters
  • Compatibility with a variety of developer environments helps ensure that you can use your tools of choice
  • Excels at analyzing a variety of data sets, including historical, real-time, structured, unstructured, and big data

In addition to driving action through data analytics, you can also create serverless compute models that trigger actions based on a particular event in IoT Hub. An earlier blog post discussed how to trigger actions using Event Grid. You can also use Azure Functions to run a snippet of code in response to events, such as uploading new files.

Azure Functions frees you up from the hassle of writing an entire app and the supporting infrastructure. Instead, all you need to do is write a short snippet of code (using your language of choice) that focuses on the problem at hand. This streamlined approach makes Functions ideal for processing orders, maintaining files, or most any other task that you need to operate on a regular basis.

Whether you're dealing with real-time or historical data, or you need to automate a function, Azure IoT has the tools to meet your needs. To learn more about working with IoT, and how easy it is to get started with your first deployment, download the IoT developer guide.